In this section we apply semantic segmentation to map the buildings over the Amsterdam region of interest.
12.1 Input data pre-processing
We intend to perform the semantic segmentation of buildings, from a Spot-7 image at 1.5m acquired over Amsterdam. The first step is to pre-process the Spot-7 image. Then, we will prepare the terrain truth that will be used for the training of our U-Net model. The main difference with the approach investigated in Part II, is that the labels are patches rather that a single label value. Hence we weed to produce a label image for our terrain truth, with the same pixel size of the Spot-7 image.
12.1.1 Satellite image pansharpening
In a sake of simplicity, we will introduce a very ...
Get Deep Learning for Remote Sensing Images with Open Source Software now with the O’Reilly learning platform.
O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.
